TWI411425B - Used in the monitoring of the eye to identify the method - Google Patents

Abstract

The present invention discloses an eye recognition/tracing method applied in personnel supervision, and an apparatus. This apparatus includes the IR illuminator, camera, filter and processing unit. The method mainly includes the pre-treatment process, the monitoring mode process and the tracing mode process. The pre-treatment process comprises the image-catching step and the image-sequencing step so as to select one picture from the vision information obtained by using IR source. The monitoring mode process consists of the step of searching candidate region, the step of analyzing candidate region and the step of confirming eye region so as to find out the eye region from the picture and analyze it. The tracing mode process is composed of the eye region tracing step and the tracing/judgment step so as to evaluate whether the eye region need to be re-positioned. By said processes and steps, this method can achieve the purpose of positioning eyes simply and certainly.

Description

Eye recognition tracking method applied to personnel monitoring

The present invention relates to an eye recognition tracking method for personnel monitoring, and more particularly to an image processing technique and apparatus for detecting and tracking an eye.

The processing unit compares the features and details of a continuous image, and obtains the required data from the image by using a preset process and analysis method, so as to be used as a part of the automatic control process, wherein The method of judging the face image of the human body has become a popular and important technology because of its wide application range. The processing and judgment method of the human face image is mainly applied to the driver's eye tracking and warning to prevent fatigue driving. The harm caused by this, of course, does not stop there. The monitoring and warning of other people can also receive automation results in conjunction with this technology, so that the importance and practicality of image tracking can be known.

The technology of the Republic of China Patent No. 436436, the method and device for continuously monitoring the driver of the vehicle, the technology is to use the mobile pixel analysis of the video homogenous two continuous video to detect the horizontal movement of the driver and to make the driver's face centered in the video. In the continuous correspondence frame, the moving pixel analysis between the continuous video frames of the homogenous video is used to detect the vertical displacement of the face of the driver in the light position to find the eye position; the conventional technology of the Republic of China Patent Publication No. 531402, eye tracking The method is that the iris information is pre-captured and the convolved iris information is convolved to find the iris position; the conventional technology of the Republic of China Patent No. I225222, the instant image detection algorithm of the instant image string, the technology thereof In the YCbCR color space that receives facial images Image data, combined motion analysis to determine the possible position of the eye; conventional technology Republic of China Patent Publication No. 200719247, the eye detection system for controlling the display device, the technology is based on the face image after the capture and the preset face data in the database Yes, to identify the position of the eye; the conventional technology of the Republic of China Patent Publication No. 200719871, is applicable to the instant face detection method of complex back, the technology is to receive the image data of the YCbCR color space of the facial image, and combine the support vector calculus for analysis and discrimination. The possible position of the eye; as can be seen from the above-mentioned various techniques, the eye seems to be the most important target for the human face, but the identification method in the conventional technique is still cumbersome, or it is necessary to capture the facial image, Iris information, etc., or need to cooperate with YCbCr technology, etc., each of the prior art has failed to provide a simple and fast eye tracking method, so that not only the higher hardware equipment can be used with the conventional technology, but also needs to be used first. The image of the person can be used, relatively speaking, its convenience and applicability are not Good; coupled with the use of various techniques to track the eyes, it is easy to be limited by weather, ambient brightness and the user's body (such as glasses, sunglasses, etc.), it is difficult to effectively achieve the effectiveness and accuracy of tracking, so the overall In terms of it, it can be said that it is quite unpractical and still needs to be improved.

The main object of the present invention is to provide an eye recognition tracking method for personnel monitoring, which can improve the conventional tracking and identification method, is too cumbersome, has high hardware limitation, is inconvenient to use, and is easily restricted to the weather and the user. Things are missing.

In order to achieve the foregoing object, the device includes an infrared ray projector, a camera, a filter, and a processing unit, and the method includes a pre-processing flow, a detection mode flow, and a tracking mode flow, wherein: the pre-processing flow includes: Take image step: take infrared image as the source of video with infrared light source; image sequence step: disassemble the captured video into several instant images for subsequent processing; detection mode process includes searching for candidate regions: Comparing the captured image with the template to select a candidate region that may be the eye; analyzing the candidate region step: confirming the eye region by using the group and geometric relationship in the candidate region; confirming the eye region step: using the real eye pattern Alignment to improve accuracy; tracking mode process includes eye area tracking step: tracking object area by object tracking method; tracking decision step: if the selected eye area cannot be tracked, the step of capturing image is restarted; if selected The eye area can be tracked, and the step of capturing the image is re-switched. Skip the search candidate area step and analyze the candidate area step; by the method described above, the infrared image is used as the video source, and can be used all day, and prevents the influence of the glasses and sunglasses provided by the user, and detects The mode flow is processed for a single screen, so the hardware processing can be speeded up, and the processing speed can be speeded up. The tracking mode process can skip the detection mode process when the tracking succeeds, which is also beneficial for the processing unit to process, so Creation is a kind of practical and progressive creation, worthy of promotion by the industry and publicity to the public.

The present invention relates to an eye recognition tracking method for personnel monitoring, wherein the device comprises an infrared light projector, a camera, a filter, and a processing unit, wherein the infrared light projector emits an infrared light source, and the camera is provided with a filter. The film is connected to the processing unit, which can filter out the light source other than the infrared light, and take the infrared image as the video for processing by the processing unit. The processing unit can be divided into an image processing unit and an image recognition unit to disassemble the video. Solve as a single screen, and then do the subsequent identification and tracking process; Continued, as shown in the first figure, this method of eye identification tracking applied to personnel monitoring includes A. Pre-processing flow, B. Detection mode flow, and C. Track the pattern process, and the meaning and detailed steps of each process are described later.

A. Pre-processing flow: The pre-processing flow is mainly achieved by using an infrared light projector and a camera with a filter. The infrared light source is used to provide an infrared light source, and the infrared light source reflected by the camera captures an infrared image, which includes a. Capture image steps and b. Image sequence step;

a. Step of capturing image: In combination with the infrared light source of the infrared light projector, the camera with the filter is used to capture the infrared image, and the infrared light projector is used to overcome the change of the brightness of the environment and the accuracy caused by the shielding of the glasses and the sunglasses. The filter is used to filter out the light source other than the infrared light, thereby improving the authenticity of the image; wherein the infrared light projector and the camera system can be disposed at any position, and the vehicle can be used as an example, and can be disposed in the cold air hole and the instrument panel. , rear view mirror, sun visor and positioning column, etc., but not limited to this.

b. Image sequence step: This step is mainly to disassemble the video from the infrared image into several instant pictures, and select one of the instant pictures, and then use the picture as the reference picture of the subsequent steps, and use the picture as the processing. In order to save system requirements and resources; the image sequence steps are organized by an image processing unit of a processing unit.

B. Detection mode flow: The detection mode process mainly processes the image obtained in the image sequence step through the image recognition unit of the processing unit, and uses c. Search for candidate area steps, d. Analysis candidate region step, e. Confirm the eye area step, and confirm the correct eye area position from the screen; c. Step of searching for candidate regions: using the processing unit to perform pattern recognition, which mainly compares the selected images with the simulated patterns, and initially selects candidate regions that are more likely to be eyes; wherein the simulation template For the description, please refer to the second and third figures. The simulation model is mainly constructed by using the pupil 11 and the iris 12 features of the human eye 10, as shown in the second figure as a normal eye. 10, wherein the portion of the pupil 11 will appear dark, and in the portion of the iris 12, the image is lighter than the pupil 11, and the pattern is constructed by using a feature as shown in the third figure. Mainly centering on the dark color, a central area 13 is established, and the surrounding area is a lighter colored surrounding area 14, and the central area 13 and the surrounding area 14 are arranged in a cross shape; therefore, the processing unit is When performing the template, the central area 13 and the surrounding area 14 are used to find the areas on the screen that match the dark color of the center and the surrounding light conditions, and select the selected areas as candidate areas. The central area 13 and the surrounding area 14 Column combination does not cross or five limited.

d. Step of analyzing candidate regions: The processing unit firstly lists a plurality of candidate regions that may be eyes through the previous step. This step further analyzes the possibility that each candidate region is an eye, and selects the region most likely to be an eye. The main method is to use Connected Component Labeling, and the grouping and geometric relationship analysis is carried out. The principle of the connected component algorithm mainly uses the selected candidate regions to use the group. The relationship between sex and geometry, such as size, shape, relative position, etc., selects the candidate area with the highest probability, taking the fourth to sixth figures as an example, wherein the fourth picture is a normal human face 20, the face The part 20 includes an eye 21 and a nose 22, and through the previous step, a light color around the center dark color is preliminarily selected, which is more likely to be a candidate area of the eye, as shown in FIG. 5, which are respectively the first candidate area. 211 and the second candidate region 221, however, are analyzed by a connected component algorithm in a group and geometric relationship, the size of the first candidate region 211 is more in line with the eye. The possibility of the size is higher, the second candidate area 221 is smaller and less likely, and the shapes of the first and second candidate areas 211 and 221 are slightly in accordance with the shape of the eye, and the relative of the first candidate area 211 If the position is moderate, the possibility is relatively high, and the relative position of the second candidate region 221 is too narrow and the possibility is low. According to the above condition analysis, the first candidate region 211 can be determined as the eye region 21, and the The eye area is positioned as shown in the sixth figure.

e. Confirmation of eye area steps: This step is mainly to make a further confirmation action, and this step can be omitted to speed up the process; the eye area step is confirmed by using the real eye pattern to improve the accuracy, compared with the simulation. The model, the real eye model refers to a pattern made by the eyes of a normal person, but does not need to limit the eyes of the user, so the accuracy can be improved, and the applicability can be considered.

C. Tracking mode process: When the detection mode process uses the processing unit to analyze and confirm the eye area from the selected image, it will enter the tracking mode process, which is processed by the image recognition unit of the processing unit to continuously track The eye area, and until the tracking mode process cannot successfully track the eye area, there is no need to enter the detection mode process, that is, in the case of successfully confirming the eye area, the pre-processing flow selects the screen and directly enters the tracking mode. Process to save the use of system resources, so this process includes f. Eye area tracking step with g. Tracking decision step; f. Eye area tracking step: When the previous process confirms the eye area from the captured image, this step uses the processing unit to track by Object Tracking. The object tracking method is numerous, and the embodiment uses tracking algorithm. The method (CamShift) is the main method, but it is not limited to this. The tracking algorithm process is as follows: I. Calculate the pixel distribution information in the entire image.

II. An initial area (the eye area in this embodiment) is selected, and a block slightly larger than this area is used as a search window, and the pixel distribution therein is calculated.

III. Perform a moving average (Mean Shift) calculation to find the center of the search window and redefine the center point and update the Zero-Moment information.

IV. Redefine the center of the search window according to the result of III, and redefine the window size according to the zero-order motion difference, and then restart II.

The tracking of the eye area is performed by the tracking algorithm, and each time the tracking is completed, the next step is performed.

g. Tracking decision step: According to the previous step, it is determined whether the tracking is successful. If the tracking fails, it indicates that the eye area has been displaced to a considerable extent. It may also be that the person has left or closed the eye, so it is necessary to return to A. The pre-processing process re-establishes the eye area. If the eye area cannot be established continuously, it can also be combined with a conventional warning mechanism to make a warning that the person has left or has closed his eyes; if the tracking is successful, return to A. Pre-processing flow and skip B. Detect mode flow and keep track of it.

The above description is only a preferred embodiment for explaining the present creation, and is not intended to impose any form limitation on the creation, so that any modification related to the creation in the same creative spirit is made. Or the changer should still be included in the scope of this creative intent.

In summary, the eye recognition identification method applied to personnel monitoring in the structure design, practical use and cost-effectiveness is indeed fully in line with the development of the industry, and the disclosed structural invention is unprecedented. Innovative structure, so its "newness" should be undoubted, and the invention can be more effective than the conventional structure, so it is also "progressive", which fully complies with the application requirements of the invention patents of the Chinese Patent Law. The stipulation is that a patent application is filed in accordance with the law, and the bureau should be reviewed at an early date and affirmed.

10‧‧‧ eyes

11‧‧‧瞳孔

12‧‧‧Iris

13‧‧‧Center

14‧‧‧ surrounding

20‧‧‧ Faces

21‧‧‧ eyes

211‧‧‧First candidate area

22‧‧‧Nose

221‧‧‧second candidate area

The first figure is a flow chart of the process of the present invention.

The second and third figures are schematic diagrams illustrating the simulation of the present invention.

The fourth to sixth figures are schematic diagrams illustrating candidate regions and eye regions of the present invention.

Claims (8)

An eye recognition tracking method applied to personnel monitoring includes the following steps: capturing an image step: capturing video with an infrared light source; and performing image sequence steps: disassembling the captured video into several instant images for subsequent processing Step of searching for a candidate region: comparing the captured image with the template to select a candidate region that may be an eye; analyzing the candidate region: confirming the eye region by using group and geometric relationships in the candidate region; eye region tracking step Tracking the eye area by object tracking method; tracking determination step: if the selected eye area cannot be tracked, the step of capturing the image is restarted; if the selected eye area can be tracked, the step of searching for the candidate area is skipped and The candidate region step is analyzed, and the read image is continuously tracked; the above steps can effectively achieve eye tracking and improve accuracy. The eye recognition tracking method applied to personnel monitoring according to Item 1 of the patent application scope, wherein each step is divided into a pre-processing process, a detection mode process, and a tracking mode process according to the attribute system; the pre-processing process includes The image capturing step and the image sequence step are performed; the detecting mode process includes a step of searching for a candidate region and a step of analyzing a candidate region; the tracking mode process includes an eye region tracking step and a tracking determination step. The eye recognition tracking method applied to personnel monitoring according to item 2 of the patent application scope, wherein the detection mode process further comprises a step of confirming an eye region after the step of analyzing the candidate region, which uses a real eye pattern for comparison To improve accuracy. An eye recognition tracking method for personnel monitoring according to the scope of claim 1 or 2 or 3, wherein the analysis candidate region step is processed by a connected component algorithm using a group geometry relationship The group and geometric relationship refers to at least one of size, shape, and relative position. The eye recognition tracking method applied to personnel monitoring according to the scope of claim 1 or 2 or 3 of the patent application, wherein the capturing image step is achieved by an infrared light projector and a camera provided with a filter. The eye recognition tracking method applied to personnel monitoring according to the first or second item or the third item of the patent application, wherein the image sequence step is achieved by an image processing unit. The eye recognition tracking method for personnel monitoring according to the first or second item or the third item of the patent application, wherein the search candidate area step, the analysis candidate area step, the eye area tracking step, and the tracking determination step It is achieved by an image recognition unit. The eye recognition tracking method applied to personnel monitoring according to item 3 of the patent application scope, wherein the step of confirming the eye area is achieved by an image recognition unit.